CN107292410A - Tunnel deformation Forecasting Methodology and device - Google Patents
Tunnel deformation Forecasting Methodology and device Download PDFInfo
- Publication number
- CN107292410A CN107292410A CN201610195345.6A CN201610195345A CN107292410A CN 107292410 A CN107292410 A CN 107292410A CN 201610195345 A CN201610195345 A CN 201610195345A CN 107292410 A CN107292410 A CN 107292410A
- Authority
- CN
- China
- Prior art keywords
- deformation
- tunnel
- matrix
- algorithm
- current time
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/11—Complex mathematical operations for solving equations, e.g. nonlinear equations, general mathematical optimization problems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Operations Research (AREA)
- Human Resources & Organizations (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Algebra (AREA)
- Software Systems (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Computing Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of tunnel deformation Forecasting Methodology and device, wherein, method includes：Obtain the deformation variable quantity prediction curve in the tunnel respectively according to the Deformation Observation data, the observation time code requirement hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm, choose hyperbola algorithm, gray system algorithm, neural network algorithm, the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum described in Kalman filtering algorithm and be used as tunnel deformation prediction curve.The tunnel deformation Forecasting Methodology that the present invention is provided employs the deformation variable quantity prediction curve that polyalgorithm obtains tunnel, by using the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum as tunnel deformation prediction curve, so as to avoid the inaccuracy that single model is brought to tunnel deformation prediction.
Description
Technical field
The present invention relates to tunnel surveying field, more particularly to a kind of tunnel deformation Forecasting Methodology and device.
Background technology
The purpose of deformation monitoring is exactly to carry out many phase repeated measures to representative deformation monitoring point, by setting up deformation model and data processing, obtains the Deformation Law of deformable body, and then make deformation analysis and prediction.Scholars have made substantial amounts of research using various Statistic analysis models to Deformation Observation data processing and prediction for many years, establish various models.
Although these deformation models have important theory significance and application value to the prediction of the fitting of Deformation Observation data and the announcement of Deformation Law and deformation, but no matter the earth's crust is settled, or building is deformed upon, its interior change mechanism is all sufficiently complex, and influenceed by many environmental factors, therefore all it is complicated stochastic system, the mechanism model for setting up deformable body is very difficult, and the feature such as uncertainty, time variation due to deformation can not be taken into account of various Statistic analysis models set up, cause that prediction error is big, precision is low.Because prediction environment is often uncertain, and be continually changing, the problem of will facing hypothetical wrong for the Individual forecast model that some factors are set up.
The content of the invention
It is an object of the invention to provide a kind of tunnel deformation Forecasting Methodology and device, to solve the problem of Individual forecast model will face hypothetical wrong in the prior art.
One aspect of the present invention provides a kind of tunnel deformation Forecasting Methodology, including：Obtain the Deformation Observation data and the corresponding observation time of the Deformation Observation data in tunnel；
Obtain the deformation variable quantity prediction curve in the tunnel respectively according to the Deformation Observation data, the observation time code requirement hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm；
The coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve in the tunnel that the specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm are obtained are obtained respectively；
By the deformation variable quantity prediction curve corresponding to the maximum of coefficient correlation described in the hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm and the minimum algorithm of prediction-error coefficients, it is defined as tunnel deformation prediction curve.
Another aspect of the present invention additionally provides a kind of tunnel deformation prediction meanss, including：Acquisition module, Deformation Observation data and the corresponding observation time of the Deformation Observation data for obtaining tunnel；
Prediction module, the deformation variable quantity prediction curve in the tunnel is obtained according to the Deformation Observation data, the observation time respectively according to specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm；
Coefficients calculation block, the coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve for obtaining the tunnel that the specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm are obtained respectively；
Choose module, for by the deformation variable quantity prediction curve corresponding to the maximum of coefficient correlation described in the hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm and the minimum algorithm of prediction-error coefficients, being defined as tunnel deformation prediction curve.
Tunnel deformation Forecasting Methodology and device that the present invention is provided, the deformation variable quantity prediction curve in tunnel is obtained by using polyalgorithm, by using the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum as tunnel deformation prediction curve, so as to avoid the inaccuracy that single model is brought to tunnel deformation prediction.
Brief description of the drawings
Fig. 1 is the flow chart of deformation Forecasting Methodology in tunnel provided in an embodiment of the present invention；
Fig. 2 is the flow chart for the tunnel deformation Forecasting Methodology that further embodiment of this invention is provided；
The flow chart for the tunnel deformation Forecasting Methodology that Fig. 3 provides for yet another embodiment of the invention；
The flow chart for the tunnel deformation Forecasting Methodology that Fig. 4 provides for another embodiment of the present invention；
The flow chart for the tunnel deformation Forecasting Methodology that Fig. 5 provides for yet another embodiment of the invention；
Fig. 6 is the structural representation of deformation prediction meanss in tunnel provided in an embodiment of the present invention；
Fig. 7 is the structural representation of deformation prediction meanss in tunnel provided in an embodiment of the present invention.
Embodiment
Embodiment one
A kind of tunnel deformation Forecasting Methodology is present embodiments provided, Fig. 1 is the flow chart of deformation Forecasting Methodology in tunnel provided in an embodiment of the present invention, as shown in figure 1, the tunnel deformation Forecasting Methodology includes：
Step 101, the Deformation Observation data and the corresponding observation time of the Deformation Observation data in tunnel are obtained.
Specifically, Deformation Observation data are multiple data, observation time certainly corresponding with Deformation Observation data is also to be multiple.
Step 102, the deformation variable quantity prediction curve in the tunnel is obtained respectively according to the Deformation Observation data, the observation time code requirement hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm.
Specifically, being predicted using polyalgorithm to the deformation variable quantity in tunnel, the prediction curve of deformation variable quantity is obtained.Typically, it is setting up the model of the deformation variable quantity in tunnel according to different algorithms, the deformation variable quantity prediction curve in tunnel is obtained according to the Deformation Observation data of input and observation time corresponding with Deformation Observation data fitting.
Step 103, the coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve in the tunnel that specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm are obtained are obtained respectively.
, it is necessary to evaluate several curves after the deformation variable quantity prediction curve in different tunnels has been got using different algorithms, it is therefore desirable to obtain the coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve in tunnel.
Step 104, hyperbola algorithm, gray system algorithm, neural network algorithm, the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum described in Kalman filtering algorithm are chosen and is used as tunnel deformation prediction curve.
Error coefficient is chosen in these prediction curves minimum, and the maximum prediction curve of coefficient correlation.Coefficient correlation refers to Deformation Observation data deformation and according to the deformation match value after the modeling of each algorithm, degree of correlation i.e. between the output valve of tunnel deformation prediction curve, if the coefficient correlation between Deformation Observation data deformation and match value is big, then illustrate that the model algorithm is capable of the deformation tendency of preferable Fitting Engineering body, the precision of prediction of deformation quantity of following phase time is high.
The tunnel deformation Forecasting Methodology that the present embodiment is provided employs the deformation variable quantity prediction curve that polyalgorithm obtains tunnel, by using the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum as tunnel deformation prediction curve, so as to avoid the inaccuracy that single model is brought to tunnel deformation prediction.
Embodiment two
The present embodiment is that above-described embodiment further explanation is illustrated, the flow chart for the tunnel deformation Forecasting Methodology that Fig. 2 provides for further embodiment of this invention, as shown in Fig. 2 the tunnel deformation Forecasting Methodology includes：
Step 201, the Deformation Observation data and the corresponding observation time of the Deformation Observation data in tunnel are obtained.
Wherein, the deformation variable quantity prediction curve for obtaining the tunnel according to the Deformation Observation data, the observation time code requirement hyperbola algorithm includes：
Step 2021, according to the initial deformation variable quantity of the observation in tunnel and final deformation variable quantity, observation time corresponding with the initial deformation variable quantity and final deformation variable quantity, hyp coefficient matrix and constant matrices are obtained using Hyperbolic Equation.
Wherein, Hyperbolic Equation is：Wherein, t is observation time, StDeformation variable quantity during for time t, S∞Final deformation variable quantity during for time t=∞, S0Deformation at initial stage variable quantity during for t=0, a is the first hyperbola parameter, and b is the second hyperbola parameter.
Specifically, obtaining the Hyperbolic Equation after deformation by Hyperbolic Equation：
(st-s0) (a+bt)=t； (2-1)
Further according to indirect adjustment formulation process, the coefficient matrix that can obtain formula (2-1) is：B is hyp coefficient matrix, turns to t=Δs t by Time Continuous t is discretei, wherein, i is the precision of discretization, and i is positive integer, can specifically be chosen according to actual needs.,For discrete time point tiThe observation deformation variable quantity in corresponding tunnel；Constant matrices is：L=[ti], L is constant matrices.
Hyperbolic Equation after the deformation typically resulted in the prior art is a+bt=t/st-s0, but such do will delete St=S0These observation data, and use the Hyperbolic Equation (s after the deformation in the present embodimentt-s0) (a+bt)=t is possible to all observation data using upper, to be more preferably predicted calculating using observing data.
Step 2022, hyp optimal first hyperbola parameter and optimal second hyperbola parameter are obtained according to hyp coefficient matrix, constant matrices and least-squares algorithm.
Specifically,Wherein, a' is optimal first hyperbola parameter, and b' is optimal second hyperbola parameter.
Step 2023, according to optimal first hyperbola parameter and optimal second hyperbola parameter, the deformation variable quantity prediction curve in tunnel is obtained.
Optimal first hyperbola parameter a' and optimal second hyperbola b' is substituted into Hyperbolic Equation, obtained
Step 203, the deformation variable quantity prediction curve in the tunnel is obtained respectively using gray system algorithm, neural network algorithm, Kalman filtering algorithm according to the Deformation Observation data, the observation time.
Wherein, step 203 can also be before step 2021.
Step 204, the coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve in the tunnel that the specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm are obtained are obtained respectively；
Step 205, hyperbola algorithm, gray system algorithm, neural network algorithm, the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum described in Kalman filtering algorithm are chosen and is used as tunnel deformation prediction curve.
The tunnel deformation Forecasting Methodology that the present embodiment is provided employs the deformation variable quantity prediction curve that polyalgorithm obtains tunnel, by using the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum as tunnel deformation prediction curve, so as to avoid the inaccuracy that single model is brought to tunnel deformation prediction.
Embodiment three
The present embodiment is to examples detailed above further explanation explanation.The flow chart for the tunnel deformation Forecasting Methodology that Fig. 3 provides for one embodiment of the invention, as shown in figure 3, the tunnel deformation Forecasting Methodology includes：
Step 301, the Deformation Observation data and the corresponding observation time of the Deformation Observation data in tunnel are obtained.
If the input quantity of neutral net is T={ tk| k=1,2, L, n }, i.e. the corresponding observation time of Deformation Observation data in tunnel；Output quantity isI.e. to the predicted value of tunnel deformation；And target reality output is X={ xk| k=1,2, L, n }, i.e. the Deformation Observation data in tunnel, k for the Deformation Observation data in tunnel quantity, specifically, neural network algorithm calculation procedure is as follows：
Step 3021, the Deformation Observation data and the corresponding observation time of the Deformation Observation data are standardized to obtain deformation standard data and observation standard time.
Deformation Observation data and the corresponding observation time of Deformation Observation data are standardized exactly the value of Deformation Observation data and the corresponding observation time of Deformation Observation data is standardized in the range of 0.1 to 0.9, this is, in order to avoid the output of two saturation regions of S type functions is minimum and maximum, and to reduce the training time.Independent variable t and dependent variableIt must all standardize (normalization).Standardize formula as follows：
Wherein, t is the corresponding observation time of Deformation Observation data, i.e. t can be t1、t2……tn。tminFor the observation time that value in t is minimum, tmaxFor the observation time that value in t is maximum.
Step 3022, hidden layer node number is determined according to the number of the Deformation Observation data and/or the corresponding observation time of the Deformation Observation data.
The problem of determination of node in hidden layer is individual sufficiently complex, experience and many experiments that often will be based on designer.Following empirical equation is often used in node in hidden layer nh determination：
In formula (3-2), ni is input layer number, i.e. ni=n equal with the quantity of the tunnel Deformation Observation data of collection；No is output layer nodes, also equal with tunnel Deformation Observation data, i.e. no=n；A is arbitrary constant, the value between [1,10].Specifically, the node number that suitable empirical equation determines hidden layer can be chosen according to different situations in formula (3-2).
Step 3023, each hidden layer node is determined to the connection weight of each output node layer and each output node layer to the initial value of the connection weight of each hidden layer node, wherein, the initial value of the connection weight is the random number in interval (- 1,1)；
Each hidden layer node is set as a small random number to the connection weight and its threshold value of the connection weight of each output node layer and each output node layer to each hidden layer node with uniform random number, it is used as the initial value of connection weight between node, interval typically takes [- 1,1].
Step 3024, the output valve of hidden layer is obtained according to observation standard time, the initial value of the connection weight of the hidden layer node to input layer, the threshold value of hidden layer node and default output function.
Wherein, hidden layer is output as：
Output layer is output as：
Specifically, hkFor hidden layer node k output,To export node layer j output, wkiFor hidden layer node k and input layer i connection weight；lwkiFor hidden layer node k threshold value；wjkFor output node layer j and hidden layer node k connection weight, lwjkTo export node layer j threshold value.Wherein i=1,2, L, ni；J=1,2, L, no；K=1,2, L, nh, wherein nh=l, l are the number of hidden layer node.
Step 3025, according to the output valve of the output layer and the deformation standard data acquisition global error, if the global error is more than predetermined threshold value, each hidden layer node is then corrected to the connection weight of each output node layer and each output node layer to the connection weight of each hidden layer node, until the global error is less than predetermined threshold value.
Wherein, global error is：
If global error is more than default threshold value, then illustrate that the forecasting inaccuracy of deformation variable quantity for tunnel is true, therefore need to correct connection weight and the connection weight of each output node layer to each hidden layer node that each hidden layer node exports node layer to each, again iteration.
Specifically, being for the connection weight adjustment formula of hidden layer node to output node layer：
It is for the connection weight adjustment formula of input layer to hidden layer node：
Wherein,Output node layer j error is represented,Hidden layer node j is represented to the correction of output node layer k connection weight, each iterative process can all correct once, η is trains speed, general η=0.01~1.
It should be noted that after the completion of prediction, predicting the outcomeIt must reduce, reduction formula is：
Wherein,For the tunnel deformation variable quantity of reduction,To export node layer j output, the i.e. predicted value to tunnel deformation variable quantity,For the maximum of tunnel deformation variable quantity predicted value,For the minimum value of tunnel deformation variable quantity.
The deformation variable quantity prediction curve in the tunnel is obtained according to the output valve of the corresponding output layer of the target global error and the output valve of the hidden layer corresponding with the output valve of the output layer.
The main thought of neural network algorithm is that learning process is divided into two stages：1st stage was forward-propagating process, i.e., calculate the output of hidden layer, output layer successively according to input layer information.2nd stage was back-propagation process, and the even reality output of output layer and the difference of desired output is more than threshold value, then adjusts connection weight.The two processes are used repeatedly so that error is tapered into, and when error reaches desired requirement, the learning process of network just terminates.
Step 3026, the deformation variable quantity prediction curve in the tunnel is obtained according to the output of the corresponding each neuron of output of the target global error.
Specifically, when global error is minimum, using the output of the output layer as the estimate of deformation quantity, so as to be worth to the prediction curve of deformation variable quantity according to the estimation of deformation quantity.
In the present embodiment, input layer is the function of time, and the connection weight weighting of connection weight and hidden layer through input layer to hidden layer to output layer obtains output layer, and output layer is the valuation of deformation quantity.Said process is repeated, is exported according to output layer and the error of desired output (deformation quantity observation) constantly adjusts connection weight, it is final to obtain the output valve for meeting computational accuracy.
Step 303, the deformation variable quantity prediction curve in the tunnel is obtained respectively according to the Deformation Observation data, the observation time code requirement hyperbola algorithm, gray system algorithm, Kalman filtering algorithm.
Step 304, the coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve in the tunnel that the specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm are obtained are obtained respectively；
Step 305, hyperbola algorithm, gray system algorithm, neural network algorithm, the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum described in Kalman filtering algorithm are chosen and is used as tunnel deformation prediction curve.
The tunnel deformation Forecasting Methodology that the present embodiment is provided employs the deformation variable quantity prediction curve that polyalgorithm obtains tunnel, by using the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum as tunnel deformation prediction curve, so as to avoid the inaccuracy that single model is brought to tunnel deformation prediction.
Example IV
The present embodiment is that examples detailed above further explanation is illustrated, the flow chart for the tunnel deformation Forecasting Methodology that Fig. 4 provides for another embodiment of the present invention, as shown in figure 4, the tunnel deformation Forecasting Methodology that the present embodiment is provided includes：
Step 401, the Deformation Observation data and the corresponding observation time of the Deformation Observation data in tunnel are obtained.
Kalman filtering algorithm is a recursive process constantly predicted, corrected, because it is when estimating state vector, substantial amounts of history observation data need not be stored, and when obtaining new observation data, the parametric filtering value that can calculate at any time newly, it is easy to handle observed result in real time, therefore Kalman filtering algorithm is increasingly being applied in dynamic monitoring data processing.The calculating process of Kalman filtering algorithm is actually constantly forecast and the process constantly corrected, therefore be easy to real-time process cycle data, and this feature exactly automatic monitoring is desired.The mathematical modeling of the algorithm includes state equation and observational equation two parts.
State parameter selection is relevant with the object (i.e. the Deformation Observation data in tunnel) and observing frequency observed.If the dynamic of the deformation variable quantity in tunnel is strong, change is fast, it is necessary to consider the rate of change and rate of acceleration of deformation variable quantity, this model is referred to as normal Fast track surgery, i.e.,：Wherein, X is the deformation variable quantity in tunnel, and kalman represents that the deformation variable quantity to tunnel carries out Kalman filtering operation；If by the dynamic of the deformation variable quantity in tunnel, strong, variation tendency is not slow, can only consider rate of change, the instantaneous variation of speed is considered as into random disturbances, this model is referred to as constant speed model, i.e.,：If change in location can be ignored in the monitored target short time completely, and observing frequency is closeer, can then deformed system be regarded as Discrete Stochastic Linear System, the instantaneous variation of position is considered as random disturbances, this model is referred to as random walk model Kalman (X).
Illustrated in the present embodiment by taking constant speed model as an example.Wherein, the deformation variable quantity prediction curve for obtaining the tunnel according to Kalman filtering algorithm according to the Deformation Observation data, the observation time includes：
Step 4021, the mathematical modeling of Kalman filtering is obtained, wherein the mathematical modeling includes：The state equation and observational equation of tunnel deformation prediction.
Specifically, tunnel deformation can be regarded as to a deformed system (hereinafter referred deformed system).
Wherein, the deformed state equation of deformed system is：
Xk=φk/k-1Xk-1+Γk/k-1Ωk-1； (4-1)
The observational equation of deformed system is：
Lk=BkXk+Δk； (4-2)
Wherein, k is observation time point, BkFor the measurement matrix at k moment, XkThe state vector matrix of etching system, φ during for kk/k-1The state-transition matrix converted for the k-1 moment to the k moment；Γk/k-1For dynamic noise matrix from the k-1 moment to k moment transformation systems；Ωk-1The dynamic noise of the system mode of etching system during for k-1；LkFor the system measurements vector at k moment, i.e. the Deformation Observation amount at k moment；ΔkFor the measurement noise at k moment.
Wherein, k-1 in the present embodiment refers to the previous moment at k moment.
Constant speed model is used in the present embodiment, i.e.,
The k-1 moment refers to k previous moment, i.e. X hereink-1It is the Deformation Observation value that the previous moment at k moment is obtained.
Step 4022, the corresponding state-transition matrix of state vector converted according to the state vector and previous moment of previous moment to current time obtains the prediction matrix of the state vector at the current time.
Specifically, Xk/k-1=φk/k-1Xk-1, wherein, φk/k-1The corresponding state-transition matrix of state vector converted for the k-1 moment to the k moment, wherein,Δtk-1,kThe observation interval between Deformation Observation value and k moment Deformation Observation values for the k-1 moment, Xk-1For the state vector at k-1 moment, Xk/k-1For the prediction matrix of the state vector at k moment.Wherein, the k-1 moment is previous moment, and the k moment is current time.State vector is exactly Deformation Observation data, Xk-1It is exactly the Deformation Observation data of previous moment, according to φk/k-1And Xk-1Obtain the predicted value of the prediction matrix, i.e. the deformation variable quantity at current time of the state vector at k moment (current time).
Step 4023, the corresponding covariance matrix of state vector at current time is obtained according to the covariance matrix of the corresponding state-transition matrix of state vector of the covariance matrix of previous moment, current time and the dynamic noise at current time.
Specifically,Wherein, φk/k-1The state-transition matrix converted for k-1 moment (previous moment) to the k moment (current time), Dk/k-1For the predicated error variance matrix to the k moment, Dk/k-1It is to be obtained according to the status predication at k-1 moment, Dk-1For the covariance matrix of k-1 moment (previous moment), the covariance matrix of previous moment refers to covariance matrix corresponding with the state vector of previous moment, Γk/k-1For the dynamic noise matrix at k moment.QkFor the covariance matrix of the dynamic noise at k moment (current time).Wherein, Qk=4 Δ t-4Rk, RkFor the covariance matrix of the observation noise at k moment (current time), wherein, R is known quantity, is the precision information of observed quantity, and Q is the error of state parameter.
It should be noted that, in step 4022 and step 4023, it is thus necessary to determine that the initial value of the state vector of initial time, i.e. state vector and covariance matrix corresponding with the state vector, specifically, the state vector X of initial time can be determined according to the Deformation Observation data in step 4010Covariance matrix D corresponding with the state vector0。
Wherein, one-dimensionalInitial value determination mode is as follows in model：X0=[x2 v0]T, wherein v0=(x2-x1)/Δt1,2,x1For first-phase tunnel deformation observation, x2For the tunnel deformation observation of the second phase,Wherein, D0For covariance matrix corresponding with the state vector of initial time, the initial value of covariance matrix is referred to as,For x2Variance,For v0And x2Covariance,For x2And v0Covariance,For v0Variance, x in this model1And x2Covariance be：
v0=(x2-x1)/Δt1,2It is that the time interval of observation time according to corresponding to first phase tunnel deformation observation and second phase tunnel deformation observation and first phase tunnel deformation observation and second phase tunnel deformation observation calculates the estimate v of an original tunnel deformation pace of change0, the initial value of deflection is x2.It is actual to filter from second phase x2Start, the deformation quantity in third phase tunnel deformation pace of change and tunnel is estimated by initial value, is gone on successively.
Step 4024, the filtering gain matrix at current time are obtained according to the corresponding covariance matrix of state vector of the observing matrix at current time, the current time and the dynamic noise variance matrix at current time.
Specifically,Wherein, JkFor the filtering gain matrix at k moment (current time), BkFor the observing matrix at k moment (current time), RkFor the covariance matrix of the observation noise at k moment (current time), RkIt should be determined according to specific monitoring means, RkDetermined by the precision of observed quantity, the deformation accuracy of measurement that different monitoring instruments (gps, spirit level, total powerstation etc.) is obtained is different.
The predicted value of the state vector of the state vector later moment in time is predicted according to the corresponding state-transition matrix of the initial value of the initial value of state vector and state vector, and obtains the covariance matrix corresponding to the state vector of later moment in time.
Step 4025, according to the filtering gain matrix at the current time, the observing matrix at current time, the observational equation amendment prediction matrix of the state vector at current time to obtain the estimation prediction matrix of the state vector at the current time.
Specifically, being the prediction matrix of the state vector according to equation below amendment current time：
Xk=Xk/k-1+Jk(Lk-BkXk/k-1), wherein, JkFor the filtering gain matrix at k moment (current time), Xk/k-1For the prediction matrix of the state vector at k moment (current time), XkThe estimation prediction matrix of state vector matrix evaluation after being corrected for the k moment (current time), the i.e. state vector at current time, LkFor the deformation observation amount at k moment (current time).
Step 4026, the corresponding covariance matrix of state vector at current time is with the corresponding estimate covariance matrix of the state vector for obtaining the current time according to the filtering gain matrix at the current time, the observing matrix amendment at the current time, until the corresponding estimate covariance matrix of state vector at the current time is less than pre-set threshold value.
According to the practical distortion amount at current time and the difference of deflection predicted value, to Dk/k-1Estimate variance covariance D is obtained after correctionk, wherein, it is that correction matrix, i.e. D are obtained according to following formulak=(I-JkBk)Dk/k-1, wherein, I is unit matrix, JkFor the filtering gain matrix at k moment (current time), BkFor the observing matrix at k moment (current time),.Specifically, the purpose for obtaining correction matrix is the predicted value X for modified state parameterk/k-1, obtain the valuation X of the predicted value of state parameterk。
Specifically, the corresponding estimate covariance matrix D of the state vector at current timekDuring less than pre-set threshold value, illustrate that the amendment of the prediction matrix of the state vector to current time in step 4024 is met and require, it is rationally accurate now to think the estimation of the deformation for tunnel.
Step 4027, the deformation variation prediction curve in the tunnel is obtained according to the corresponding estimate covariance Output matrix of estimation prediction matrix and the state vector at the current time of the state vector at the current time.
Specifically, when the corresponding estimate covariance matrix pre-set threshold value of the state vector at current time, the estimation prediction matrix of the state vector at corresponding current time just meets the prediction for tunnel deformation, therefore to the corresponding estimate covariance Output matrix of state vector at current time, the deformation variation prediction curve in tunnel is obtained.
Specifically, when performing the present embodiment, can obtain the m phases observe data, progress Kalman filtering process.
Kalman filtering process includes：
The tentative prediction of state parameter：Xk/k-1=φk/k-1Xk-1；
State covariance battle array prediction：
Filtering gain matrix：
Amendment is filtered to state parameter：Xk=Xk/k-1+Jk(Lk-BkXk/k-1)；
State covariance battle array estimation：Dk=(I-JkBk)Dk/k-1。
If obtaining new observation, remove downpayment observation, new observation reconstituted into the m phases as eventually observation observes data, step 4022 is re-executed to step 4025, to reach the purpose of dynamic filter.
Step 403, the deformation variable quantity prediction curve in the tunnel is obtained respectively according to the Deformation Observation data, the observation time code requirement hyperbola algorithm, gray system algorithm, neural network algorithm.
Wherein, step 403 can also be before step 4021.
Step 404, the coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve in the tunnel that the specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm are obtained are obtained respectively；
Step 405, hyperbola algorithm, gray system algorithm, neural network algorithm, the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum described in Kalman filtering algorithm are chosen and is used as tunnel deformation prediction curve.
The tunnel deformation Forecasting Methodology that the present embodiment is provided employs the deformation variable quantity prediction curve that polyalgorithm obtains tunnel, by using the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum as tunnel deformation prediction curve, so as to avoid the inaccuracy that single model is brought to tunnel deformation prediction.
Embodiment five
The present embodiment is to the further supplementary notes of above-described embodiment.The flow chart for the tunnel deformation Forecasting Methodology that Fig. 5 provides for yet another embodiment of the invention, as shown in figure 5, the tunnel deformation Forecasting Methodology that the present embodiment is provided includes：
Step 501, the Deformation Observation data and the corresponding observation time of the Deformation Observation data in tunnel are obtained.
Due to gray system algorithm model GM (1,1) it is with constant duration Series Modeling, it requires the data break used for even time interval, and the settlement observation data in real work are typically unequal time-interval, to solve this contradiction, it is isochronous sequence to be applied unequal time-interval sedimentation sequence transitions before Accumulating generation and after inverse accumulated generating.Specifically, the deformation variable quantity prediction curve for obtaining the tunnel according to gray system algorithm according to the Deformation Observation data, the observation time includes：
Step 5011, the Deformation Observation data in the tunnel are subjected to cubic spline difference and obtain even time interval deformation data.
It is acquired specifically, the Deformation Observation data in tunnel are not constant durations, i.e., the Deformation Observation data in tunnel are non-equal time interval deformation variable quantity, therefore, is even time interval deformation data by the non-equal time interval deformation variable quantity sequence transformation.
Wherein, non-equal time interval deformation variable quantity sequence is designated as：
X1 (0)=[x1 (0)(t1),x1 (0)(t2),L,x1 (0)(tk),L,x1 (0)(tn)]；
Wherein, x1 (0)(n) it is n-th of deformation variable quantity.Wherein, in even time interval deformation data, average time interval is：
Wherein, Δ t0For average time interval.T is observation time corresponding with deformation variable quantity, and i is time point sequence number, i=1,2, K, n.
When obtaining even time interval deformation data, the main cubic spline interpolation algorithm using First Boundary Condition.
Wherein, the spline coefficients in cubic spline interpolation are：
sk,0=yk； (5-2)
Wherein, sk,0For the first spline coefficients, sk,1For the second spline coefficients, sk,2For the 3rd spline coefficients, sk,3For the 4th spline coefficients.mk=S "k(xk), hk=xi-xi-1, wherein Si(x) it is [xi-1,xi] on spline function, x hereini=ti, ykRepresent tunnel deformation amount.
Time point to be interpolated is substituted into formula (5-2) to formula (5-5) spline function, even time interval deformation data X is obtained2 (0)={ x2 (0)(ti) | i=1,2, L n }.Wherein, x2 (0)(ti) be even time interval deformation data element.
Step 5012, cumulative is done to the element in even time interval deformation data, obtains formation sequence.
Wherein, formation sequence is：X2 (1)={ x2 (1)(ti) | i=1,2, L n } X2 (1)={ x2 (1)(ti) | i=1,2, L n }, wherein,X2 (1)For formation sequence, x2 (1)(ti) be formation sequence element.
Step 5013, coefficient matrix and constant matrices are obtained according to the element of formation sequence.
Wherein, coefficient matrix B and constant matrices L are respectively：
L=[x2 (0)(t2) x2 (0)(t2) L x2 (0)(tn)]Τ； (5-7)
Step 5014, the first parameter and the second parameter of gray system are obtained using least square method according to coefficient matrix and constant matrices；
Wherein, because the deformation quantity in tunnel can approximately regard it as time t function, the differential equation of first order of the deformation quantity in tunnel can be designated as
It is further gray system can be obtained according to least square method the first parameter and the second parameter be：
Wherein, a is the first parameter, and b is the second parameter.
Step 5015, the deformation variable quantity prediction curve in the tunnel is obtained according to first parameter and the second parameter.
Specifically, the deformation variable quantity prediction curve in tunnel is：
Wherein, tiFor i-th of observation time point,For tiThe predicted value of the deformation variable quantity in the tunnel at moment, Δ toFor average time interval.
Further by the regressive reduction-type of formula (5-9)It is availableThe predicted value of sequence is：
As t=∞,Calculated value is equal to limiting value a/b, and the value may be considered in the final deformation quantity of monitoring point, program implement, and final settlement is a/b.
Step 503, the deformation variable quantity prediction curve in the tunnel is obtained respectively according to the Deformation Observation data, the observation time code requirement hyperbola algorithm, Kalman filtering algorithm, neural network algorithm.
Wherein, step 503 can also be before step 5021.
Step 504, the coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve in the tunnel that the specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm are obtained are obtained respectively；
Step 505, hyperbola algorithm, gray system algorithm, neural network algorithm, the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum described in Kalman filtering algorithm are chosen and is used as tunnel deformation prediction curve.
The tunnel deformation Forecasting Methodology that the present embodiment is provided employs the deformation variable quantity prediction curve that polyalgorithm obtains tunnel, by using the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum as tunnel deformation prediction curve, so as to avoid the inaccuracy that single model is brought to tunnel deformation prediction.
Embodiment six
The present embodiment is that the general principle of the cubic spline interpolation in embodiment five is explained.
The general principle of cubic spline functions is as follows：
IfThere are N+1 point, xkFor the time of kth point, ykFor the deformation quantity of kth point, k=0,1 ... N, wherein：
A=x0<x1<L<xN=b； (6-1)
A and b is the boundary parameter of cubic spline functions.If there is cubic polynomial Sk(x), coefficient is Sk,0,Sk,1,Sk,2,Sk,3Meeting formula (6-2), then S (x) is called cubic spline function to the property of formula (6-6).
Specifically, formula (6-2) to formula (6-6) is：
Sk(x)=sk,0+sk,1(x-xk)+sk,2(x-xk)2+sk,3(x-xk)3； (6-2)
S(xk)=yk； (6-3)
Sk(xk+1)=Sk+1(xk+1)； (6-4)
S'k(xk+1)=S'k+1(xk+1)； (6-5)
S″k(xk+1)=S "k+1(xk+1)； (6-6)
Wherein, in formula (6-3), ykFor kth point deformation quantity, k span is 0,1 ..., N；In formula (6-4) into (6-6), k span is 0,1 ..., N-2.
Because S (x) is segmental cubic polynomials, S (x) second dervative is in interval [x0,xN] in be piecewise linearity.According to linear lagrange interpolation S " (x)=Sk" (x) can be expressed as：
In order to simplify formula (6-7), m is usedk=S " (xk)、mk+1=S (xk+1) and hk=xk+1-xkFormula (6-7) is substituted into, wherein,
Formula (5-9) is integrated twice, therefore two integral constants can be introduced, following formula is can obtain：
Wherein, pkFor first integral constant, qkFor second integral constant.
By xkAnd xk+1Formula (5-10) is substituted into, and utilizes yk=Sk(xk) and yk+1=Sk(xk+1) two equations can be obtained：
Solve pkAnd qk, and bring the result of gained into equation (6-10)：
Derived function is solved to formula (6-13), can be obtained：
hk-1mk-1+2(hk-1+hk)mk+hkmk+1=uk； (6-14)
Wherein, uk=6 (dk-dk-1), k=1 ..., N-1, dk=(yk+1-yk)/hk。
Equation below can be obtained by formula (6-14)：
2(h0+h1)m1+h1m2=u1-h0m0； (6-15)
hk-1mk-1+2(hk-1+hk)mk+hkmk+1=uk； (6-16)
hN-1mN-2+2(hN-2+hN-1)mN-1=uN-1-hN-1mN； (6-17)
Formula (6-15), (6-16), (6-17) are recombinated, triangular linear equation group HM=V is obtained, is expressed as
Formula (6-18) has Strictly diagonal dominance, calculates coefficient { mkAfter S can be calculated by equation belowk(x) spline coefficients { sk,j}。
sk,0=yk；
According to Least Square Method the first parameter a and the second parameter b.
Embodiment seven
The present embodiment is to above-described embodiment further explanation explanation.Wherein, the coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve in the tunnel that the specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm are obtained are obtained respectively, including：
The calculation formula of wherein coefficient correlation is：
Wherein, r is coefficient correlation, xiFor the original Deformation Observation value in tunnel, yiThe prediction predicted value of the deformation variable quantity in the tunnel obtained according to the algorithm of above-described embodiment is represented, i value is 1 to n, and n is the quantity that deformation variable quantity in tunnel is predicted, value is positive integer.
Prediction-error coefficients are：
Wherein, W is prediction-error coefficients.
Embodiment eight
A kind of tunnel deformation prediction meanss are present embodiments provided, for performing the tunnel deformation Forecasting Methodology in above-described embodiment, the wherein device can be fixed terminal, such as computer, naturally it is also possible to be handheld terminal.Fig. 6 is the structural representation of deformation prediction meanss in tunnel provided in an embodiment of the present invention, as shown in fig. 6, the device includes：Acquisition module 801, prediction module 802, coefficients calculation block 803 and selection module 804.
Wherein, acquisition module 801, Deformation Observation data and the corresponding observation time of the Deformation Observation data for obtaining tunnel；
Prediction module 802, which distinguishes acquisition module 801, coefficients calculation block 803 and chooses module 80, to be connected, and prediction module 802 is used for the deformation variable quantity prediction curve for obtaining the tunnel respectively according to specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm according to the Deformation Observation data, the observation time；
Coefficients calculation block 803 is connected with prediction module 802, the coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve for obtaining the tunnel that the specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm are obtained respectively；
Module 804 is chosen to be connected with coefficients calculation block 803 and prediction module 802, for by the deformation variable quantity prediction curve corresponding to the maximum of coefficient correlation described in the hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm and the minimum algorithm of prediction-error coefficients, being defined as tunnel deformation prediction curve.
The tunnel deformation prediction meanss that the present embodiment is provided can obtain the deformation variable quantity prediction curve in tunnel using polyalgorithm, by using the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum as tunnel deformation prediction curve, so as to avoid the inaccuracy that single model is brought to tunnel deformation prediction.
Embodiment nine
The present embodiment is that explanation is further explained to embodiment eight, and Fig. 7 is the structural representation of deformation prediction meanss in tunnel provided in an embodiment of the present invention, as shown in fig. 7, the device includes：Acquisition module 901, prediction module 902, coefficients calculation block 903 and selection module 904.
Wherein, prediction module 902 includes：Specification hyperbola algoritic module 9021, gray system algoritic module 9022, neural network algorithm module 9023 and Kalman filtering algorithm module 9024.
Wherein, specification hyperbola algoritic module 9021 is connected with acquisition module 901 and coefficients calculation block 903 respectively, also it is connected with choosing module 904, specification hyperbola algoritic module 9021 is used for initial deformation variable quantity, final deformation variable quantity, the initial deformation variable quantity and the corresponding observation time of final deformation variable quantity of the observation according to tunnel, and hyp coefficient matrix and constant matrices are obtained using Hyperbolic Equation；
Specification hyperbola algoritic module 9021 is additionally operable to obtain hyp optimal first hyperbola parameter and optimal second hyperbola parameter according to the hyp coefficient matrix, the constant matrices and least-squares algorithm；
Specification hyperbola algoritic module 9021 is additionally operable to the deformation variable quantity prediction curve that the specification hyperbola algoritic module is additionally operable to obtain the tunnel according to optimal first hyperbola parameter and optimal second hyperbola parameter.
Gray system algoritic module 9022 is connected with acquisition module 901 and coefficients calculation block 903 respectively, also it is connected with choosing module 904, gray system algoritic module 9022 is used to the Deformation Observation data in the tunnel carrying out cubic spline difference acquisition even time interval deformation data；
Gray system algoritic module 9022 is additionally operable to do cumulative to the element in the even time interval deformation data, obtains formation sequence；Coefficient matrix and constant matrices are obtained according to the element of formation sequence；
Gray system algoritic module 9022 is additionally operable to obtain the first parameter and the second parameter of gray system using least square method according to coefficient matrix and constant matrices；
Gray system algoritic module 9022 is additionally operable to obtain the deformation variable quantity prediction curve in the tunnel according to first parameter and the second parameter.
Neural network algorithm module 9023 is connected with acquisition module 901 and coefficients calculation block 903 respectively, also it is connected with choosing module 904, neural network algorithm module 9023 is used to standardize the Deformation Observation data and the corresponding observation time of the Deformation Observation data to obtain deformation standard data and observation standard time.
Neural network algorithm module 9023 is additionally operable to determine hidden layer node number according to the number of the Deformation Observation data and/or the corresponding observation time of the Deformation Observation data.
Neural network algorithm module 9023 is additionally operable to determine each hidden layer node to the connection weight and the initial value of each output node layer to the connection weight of each hidden layer node of each output node layer, wherein, the initial value of the connection weight is the random number in interval (- 1,1).
Neural network algorithm module 9023 is additionally operable to obtain the output valve of hidden layer according to observation standard time, the initial value of the connection weight of the hidden layer node to input layer, the threshold value of hidden layer node and default output function；
Output valve, the initial value of the output node layer to the connection weight of hidden layer node, the threshold value for exporting node layer and the default output function that neural network algorithm module 9023 is additionally operable to according to the hidden layer obtain the output valve of output layer.
Neural network algorithm module 9023 is additionally operable to output valve and the deformation standard data acquisition global error according to the output layer, if the global error is more than predetermined threshold value, each hidden layer node is then corrected to the connection weight of each output node layer and each output node layer to the connection weight of each hidden layer node, until the global error is less than predetermined threshold value.
Neural network algorithm module 9023 is additionally operable to obtain the deformation variable quantity prediction curve in the tunnel according to the output valve of the corresponding output layer of the target global error and the output valve of the hidden layer corresponding with the output valve of the output layer.
Kalman filtering algorithm module 9024 is connected with acquisition module 901 and coefficients calculation block 903 respectively, is also connected with choosing module 904, and Kalman filtering algorithm module 9024 is used for the mathematical modeling for obtaining Kalman filtering, wherein the mathematical modeling includes：The state equation and observational equation of tunnel deformation prediction；
Kalman filtering algorithm module 9024 is additionally operable to the prediction matrix of the state vector at the state vector corresponding state-transition matrix acquisition current time converted according to the state vector and previous moment of previous moment to current time；
The covariance matrix that Kalman filtering algorithm module 9024 is additionally operable to covariance matrix according to previous moment, the corresponding state-transition matrix of the state vector at current time and the dynamic noise at current time obtains the corresponding covariance matrix of state vector at current time；
Observing matrix, the corresponding covariance matrix of the state vector at the current time and the dynamic noise variance matrix at current time that Kalman filtering algorithm module 9024 was additionally operable to according to current time obtain the filtering gain matrix at current time；
Kalman filtering algorithm module 9024 is additionally operable to filtering gain matrix according to the current time, the observing matrix at current time, the prediction matrix of the state vector at current time described in the observational equation amendment to obtain the estimation prediction matrix of the state vector at the current time；
Kalman filtering algorithm module 9024 is additionally operable to the corresponding covariance matrix of state vector of filtering gain matrix according to the current time, the observing matrix at the current time and current time described in the corresponding covariance matrix amendment of the state vector at the current time, until the second correction matrix of the corresponding covariance matrix of state vector at the current time is less than pre-set threshold value；
Kalman filtering algorithm module 9024 is additionally operable to obtain the deformation variation prediction curve in the tunnel according to the estimation prediction matrix and the corresponding estimate covariance Output matrix of the state vector at the current time of the state vector at the current time.
It should be noted that, acquisition module 901 in the present embodiment is connected with specification hyperbola algoritic module 9021, gray system algoritic module 9022, neural network algorithm module 9023 and Kalman filtering algorithm module 9024 respectively, but the acquisition module 901 that draws merely exemplary in Fig. 7 is connected with specification hyperbola algoritic module 9021.
The tunnel deformation prediction meanss that the present embodiment is provided can obtain the deformation variable quantity prediction curve in tunnel using polyalgorithm, by using the deformation variable quantity prediction curve corresponding to the algorithm that coefficient correlation is maximum and prediction-error coefficients are minimum as tunnel deformation prediction curve, so as to avoid the inaccuracy that single model is brought to tunnel deformation prediction.
Finally it should be noted that：The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although the present invention is described in detail with reference to the foregoing embodiments, the ordinary skill passenger of this area should be understood：It can still modify to the technical scheme described in foregoing embodiments, or carry out equivalent substitution to which part technical characteristic；And these modifications or replacement, the essence of appropriate technical solution is departed from the spirit and scope of various embodiments of the present invention technical scheme.
Claims (10)
1. a kind of tunnel deformation Forecasting Methodology, it is characterised in that including：
Obtain the Deformation Observation data and the corresponding observation time of the Deformation Observation data in tunnel；
According to the Deformation Observation data, the observation time code requirement hyperbola algorithm, gray system
Algorithm, neural network algorithm, Kalman filtering algorithm obtain the deformation variable quantity prediction in the tunnel respectively
Curve；
The specification hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman are obtained respectively
The coefficient correlation and prediction-error coefficients of the deformation variable quantity prediction curve in the tunnel that filtering algorithm is obtained；
By in the hyperbola algorithm, gray system algorithm, neural network algorithm, Kalman filtering algorithm
Deformation variable quantity prediction corresponding to the algorithm that the coefficient correlation is maximum and prediction-error coefficients are minimum is bent
Line, is defined as tunnel deformation prediction curve.
2. according to the method described in claim 1, it is characterised in that according to the Deformation Observation data,
The observation time, code requirement hyperbola algorithm obtains the deformation variable quantity prediction curve bag in the tunnel
Include：
Become according to the initial deformation variable quantity of the observation in tunnel, final deformation variable quantity, the initial deformation
Change amount and the corresponding observation time of final deformation variable quantity, hyp coefficient is obtained using Hyperbolic Equation
Matrix and constant matrices；
Hyperbola is obtained according to the hyp coefficient matrix, the constant matrices and least-squares algorithm
Optimal first hyperbola parameter and optimal second hyperbola parameter；
According to optimal first hyperbola parameter and optimal second hyperbola parameter, the tunnel is obtained
Deformation variable quantity prediction curve.
3. according to the method described in claim 1, it is characterised in that according to the Deformation Observation data,
The observation time, the deformation variable quantity prediction curve in the tunnel, bag are obtained using gray system algorithm
Include：
The Deformation Observation data in the tunnel are subjected to cubic spline difference and obtain even time interval deformation time sequence
Row；
Cumulative is done to the element in the even time interval deformation data, formation sequence is obtained；
Coefficient matrix and constant matrices are obtained according to the element of formation sequence；
The first parameter and the of gray system is obtained using least square method according to coefficient matrix and constant matrices
Two parameters；
The deformation variable quantity prediction curve in the tunnel is obtained according to first parameter and the second parameter.
4. according to the method described in claim 1, it is characterised in that according to the Deformation Observation data,
The observation time, the deformation variable quantity prediction curve in the tunnel, bag are obtained using neural network algorithm
Include：
The Deformation Observation data and the corresponding observation time of the Deformation Observation data are standardized to obtain
Take deformation standard data and observation standard time；
According to the Deformation Observation data and/or the number of the corresponding observation time of the Deformation Observation data
Determine hidden layer node number；
Determine that each hidden layer node is arrived to the connection weight of each output node layer and each output node layer
The initial value of the connection weight of each hidden layer node, wherein, the initial value of the connection weight is interval (- 1,1)
Interior random number；
According to observation the standard time, the initial value of the connection weight of the hidden layer node to input layer,
The threshold value of hidden layer node and default output function obtain the output valve of hidden layer；
According to the output valve of the hidden layer, the output node layer to the connection weight of hidden layer node
Initial value, the threshold value for exporting node layer and the default output function obtain the output valve of output layer；
According to the output valve of the output layer and the deformation standard data acquisition global error, if described complete
Office's error is more than predetermined threshold value, then corrects each hidden layer node to the connection weight of each output node layer
Node layer is exported to the connection weight of each hidden layer node with each, until the global error is less than in advance
If threshold value；
According to the output valve of the corresponding output layer of the target global error and the output valve with the output layer
The output valve of the corresponding hidden layer obtains the deformation variable quantity prediction curve in the tunnel.
5. according to the method described in claim 1, it is characterised in that according to the Deformation Observation data,
The observation time, the deformation variable quantity prediction curve in the tunnel is obtained using Kalman filtering algorithm,
Including：
The mathematical modeling of Kalman filtering is obtained, wherein the mathematical modeling includes：Tunnel deformation prediction
State equation and observational equation；
The state vector correspondence converted according to the state vector and previous moment of previous moment to current time
State-transition matrix obtain current time state vector prediction matrix；
According to the corresponding state-transition matrix of the state vector of the covariance matrix of previous moment, current time
And the covariance matrix of the dynamic noise at current time obtains the corresponding association side of state vector at current time
Poor matrix；
According to the corresponding covariance matrix of the state vector of the observing matrix at current time, the current time
The filtering gain matrix at current time are obtained with the dynamic noise variance matrix at current time；
According to the filtering gain matrix at the current time, the observing matrix at current time, the observation side
The prediction matrix of the state vector at current time described in Cheng Xiuzheng is to obtain the state vector at the current time
Estimation prediction matrix；
According to the filtering gain matrix at the current time, the observing matrix amendment at the current time
The corresponding covariance matrix of state vector at current time is corresponding with the state vector for obtaining the current time
Estimate covariance matrix, until the current time the corresponding estimate covariance matrix of state vector it is small
In pre-set threshold value；
According to the state vector at the current time estimation prediction matrix and the state at the current time to
Measure the deformation variation prediction curve that corresponding estimate covariance Output matrix obtains the tunnel.
6. a kind of tunnel deformation prediction meanss, it is characterised in that including：
Acquisition module, the corresponding sight of Deformation Observation data and the Deformation Observation data for obtaining tunnel
The survey time；
Prediction module, according to the Deformation Observation data, the observation time according to specification hyperbola algorithm,
The deformation that gray system algorithm, neural network algorithm, Kalman filtering algorithm obtain the tunnel respectively becomes
Change amount prediction curve；
Coefficients calculation block, for obtaining the specification hyperbola algorithm, gray system algorithm, god respectively
The coefficient correlation of the deformation variable quantity prediction curve in the tunnel obtained through network algorithm, Kalman filtering algorithm
And prediction-error coefficients；
Module is chosen, for by the hyperbola algorithm, gray system algorithm, neural network algorithm, card
The shape corresponding to algorithm that coefficient correlation described in Kalman Filtering algorithm is maximum and prediction-error coefficients are minimum
Become variable quantity prediction curve, be defined as tunnel deformation prediction curve.
7. device according to claim 1, it is characterised in that the prediction module includes：
Specification hyperbola algoritic module, for the initial deformation variable quantity of the observation according to tunnel, most end form
Become variable quantity, the initial deformation variable quantity and the corresponding observation time of final deformation variable quantity, using double
Curvilinear equation obtains hyp coefficient matrix and constant matrices；
It is additionally operable to be obtained according to the hyp coefficient matrix, the constant matrices and least-squares algorithm
Hyp optimal first hyperbola parameter and optimal second hyperbola parameter；
The specification hyperbola algoritic module is additionally operable to be additionally operable to according to optimal first hyperbola parameter and optimal
Second hyperbola parameter, obtains the deformation variable quantity prediction curve in the tunnel.
8. device according to claim 1, it is characterised in that the prediction module also includes：
Gray system algoritic module, for the Deformation Observation data in the tunnel to be carried out into cubic spline difference
Obtain even time interval deformation data；
It is additionally operable to do cumulative to the element in the even time interval deformation data, obtains formation sequence；Root
Coefficient matrix and constant matrices are obtained according to the element of formation sequence；
It is additionally operable to obtain the first ginseng of gray system using least square method according to coefficient matrix and constant matrices
Number and the second parameter；
It is additionally operable to predict song according to the deformation variable quantity that first parameter and the second parameter obtain the tunnel
Line.
9. device according to claim 1, it is characterised in that the prediction module also includes：
Neural network algorithm module, for by the Deformation Observation data and the Deformation Observation data pair
The observation time answered standardizes to obtain deformation standard data and observation standard time；
It is additionally operable to according to the Deformation Observation data and/or the corresponding observation time of the Deformation Observation data
Number determine hidden layer node number；
It is additionally operable to determine each hidden layer node to the connection weight and each output layer of each output node layer
Node to the connection weight of each hidden layer node initial value, wherein, the initial value of the connection weight is area
Between random number in (- 1,1)；
It is additionally operable to according to observation standard time, the connection weight of the hidden layer node to input layer
Initial value, the threshold value of hidden layer node and default output function obtain the output valve of hidden layer；
It is additionally operable to output valve according to the hidden layer, the connection of the output node layer to hidden layer node
The initial value of weights, the threshold value for exporting node layer and the default output function obtain the output valve of output layer；
The output valve and the deformation standard data acquisition global error according to the output layer are additionally operable to, if
The global error is more than predetermined threshold value, then corrects each hidden layer node to the company of each output node layer
Weights and each output node layer are connect to the connection weight of each hidden layer node, until the global error
Less than predetermined threshold value；
Be additionally operable to according to the output valve of the corresponding output layer of the target global error and with the output layer
The output valve of the corresponding hidden layer of output valve obtains the deformation variable quantity prediction curve in the tunnel.
10. device according to claim 1, it is characterised in that the prediction module also includes：
Kalman filtering module, the mathematical modeling for obtaining Kalman filtering, wherein the mathematical modeling
Including：The state equation and observational equation of tunnel deformation prediction；
Be additionally operable to state from previous moment to current time that converted according to the state vector and previous moment of to
Measure the prediction matrix that corresponding state-transition matrix obtains the state vector at the current time；
The covariance matrix according to previous moment, the corresponding state of the state vector at current time is additionally operable to turn
Move the state vector correspondence at the covariance matrix acquisition current time of matrix and the dynamic noise at current time
Covariance matrix；
It is additionally operable to observing matrix according to current time, the corresponding association side of the state vector at the current time
Poor matrix and the dynamic noise variance matrix at current time obtain the filtering gain matrix at current time；
Be additionally operable to filtering gain matrix according to the current time, it is the observing matrix at current time, described
The prediction matrix of the state vector at current time described in observational equation amendment is to obtain the shape at the current time
The estimation prediction matrix of state vector；
Be additionally operable to filtering gain matrix according to the current time, the observing matrix at the current time and
The state vector pair at current time described in the corresponding covariance matrix amendment of state vector at the current time
The covariance matrix answered, until the second of the corresponding covariance matrix of state vector at the current time repaiies
Positive matrices is less than pre-set threshold value；
It is additionally operable to according to the estimation prediction matrix of the state vector at the current time and the current time
The corresponding estimate covariance Output matrix of state vector obtains the deformation variation prediction curve in the tunnel.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610195345.6A CN107292410A (en) | 2016-03-30 | 2016-03-30 | Tunnel deformation Forecasting Methodology and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610195345.6A CN107292410A (en) | 2016-03-30 | 2016-03-30 | Tunnel deformation Forecasting Methodology and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107292410A true CN107292410A (en) | 2017-10-24 |
Family
ID=60086767
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610195345.6A Pending CN107292410A (en) | 2016-03-30 | 2016-03-30 | Tunnel deformation Forecasting Methodology and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107292410A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108332649A (en) * | 2018-02-07 | 2018-07-27 | 桂林电子科技大学 | A kind of landslide deformation comprehensive pre-warning method and system |
| CN110333494A (en) * | 2019-04-10 | 2019-10-15 | 马培峰 | A kind of InSAR timing deformation prediction method, system and relevant apparatus |
| CN112015778A (en) * | 2020-08-19 | 2020-12-01 | 上海满盛信息技术有限公司 | Water fingerprint prediction algorithm |
| CN112307414A (en) * | 2020-10-29 | 2021-02-02 | 上海勘察设计研究院(集团)有限公司 | Tunnel scanning point cloud resolving method based on nonlinear Kalman filtering |
| CN113642812A (en) * | 2021-10-15 | 2021-11-12 | 西南交通大学 | Beidou-based micro-deformation prediction method, device, equipment and readable storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103885386A (en) * | 2014-04-08 | 2014-06-25 | 北京工业大学 | Gray model thermal error data processing method based on Kalman filtering |
| CN105160423A (en) * | 2015-09-14 | 2015-12-16 | 河海大学常州校区 | Photovoltaic power generation prediction method based on Markov residual error correction |
-
2016
- 2016-03-30 CN CN201610195345.6A patent/CN107292410A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103885386A (en) * | 2014-04-08 | 2014-06-25 | 北京工业大学 | Gray model thermal error data processing method based on Kalman filtering |
| CN105160423A (en) * | 2015-09-14 | 2015-12-16 | 河海大学常州校区 | Photovoltaic power generation prediction method based on Markov residual error correction |
Non-Patent Citations (1)
| Title |
|---|
| "冻土地区高速铁路软土路基沉降预测模型研究": "邓锐等", 《铁道勘察》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108332649A (en) * | 2018-02-07 | 2018-07-27 | 桂林电子科技大学 | A kind of landslide deformation comprehensive pre-warning method and system |
| CN108332649B (en) * | 2018-02-07 | 2020-04-24 | 桂林电子科技大学 | Landslide deformation comprehensive early warning method and system |
| CN110333494A (en) * | 2019-04-10 | 2019-10-15 | 马培峰 | A kind of InSAR timing deformation prediction method, system and relevant apparatus |
| CN112015778A (en) * | 2020-08-19 | 2020-12-01 | 上海满盛信息技术有限公司 | Water fingerprint prediction algorithm |
| CN112307414A (en) * | 2020-10-29 | 2021-02-02 | 上海勘察设计研究院(集团)有限公司 | Tunnel scanning point cloud resolving method based on nonlinear Kalman filtering |
| CN112307414B (en) * | 2020-10-29 | 2022-07-29 | 上海勘察设计研究院(集团)有限公司 | Tunnel scanning point cloud resolving method based on nonlinear Kalman filtering |
| CN113642812A (en) * | 2021-10-15 | 2021-11-12 | 西南交通大学 | Beidou-based micro-deformation prediction method, device, equipment and readable storage medium |
| CN113642812B (en) * | 2021-10-15 | 2022-02-08 | 西南交通大学 | Beidou-based micro-deformation prediction method, device, equipment and readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107292410A (en) | Tunnel deformation Forecasting Methodology and device | |
| CN107765347B (en) | Short-term wind speed prediction method based on Gaussian process regression and particle filtering | |
| Madhukar et al. | State estimation using extended kalman filter and unscented kalman filter | |
| CN104392136B (en) | A kind of high accuracy data fusion method towards high dynamic LDPC code robust measure | |
| CN108682023A (en) | Close coupling Unscented kalman tracking filter algorithm based on Elman neural networks | |
| CN108764568B (en) | Data prediction model tuning method and device based on LSTM network | |
| CN110309481B (en) | Model modeling method for farmland soil reaction dynamic process | |
| CN109388778A (en) | A kind of iteration volume point Unscented kalman filtering method | |
| CN110443724A (en) | A kind of electric system fast state estimation method based on deep learning | |
| CN112043258A (en) | Dynamic heart rate prediction method, device, equipment and readable storage medium | |
| Marti et al. | Adaptive stochastic path planning for robots–real-time optimization by means of neural networks | |
| CN116595897A (en) | Nonlinear dynamic system state estimation method and device based on message passing | |
| CN117313029A (en) | Multi-sensor data fusion method based on Kalman filtering parameter extraction and state updating | |
| CN114239796A (en) | Power system state estimation method based on extended Kalman filtering | |
| CN108804721B (en) | Oil pumping machine fault diagnosis method based on self-adaptive unscented Kalman filtering and RBF neural network | |
| CN113625182A (en) | Method for on-line estimating battery state | |
| Rivero et al. | Short-term rainfall time series prediction with incomplete data | |
| CN103914628B (en) | A kind of Space teleoperation system output state Forecasting Methodology | |
| CN110366232B (en) | Sensor transmission energy control method for remote state estimation | |
| CN115865702A (en) | Distributed fusion estimation method with data attenuation under network scheduling strategy | |
| CN107145474B (en) | Kalman filtering algorithm based on autoregressive model | |
| CN109884893B (en) | Multi-process variable dynamic time lag estimation method | |
| CN106814608A (en) | PREDICTIVE CONTROL adaptive filter algorithm based on Posterior probability distribution | |
| CN104899467A (en) | Frequency shift correction method and system based on frequency control | |
| CN117131465B (en) | Single-pipe tower damage identification method and device, electronic equipment and readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171024 |
|
| RJ01 | Rejection of invention patent application after publication |